TPU used that horrific test method (fully unlocked FPS) and those special gameworks/physX titles where nV gets huge boost but not many people play them. So, if you exclude fps above 144fps and take out benchmarketing titles. You'll see different scores. Btw, Hilbert, can you do weight averages as summarization? You know, now everyone has his sets of games for reviews and adds new ones in time/removes old ones. And average for all is just plain average. What about weight average based on total play time on steam or active player base in last 3/6 months? That way you get comparison of what matters to people. Not that your set is bad or anything, but you may set new trend. And show other sites how poor selection of games they use.